MALDI Biotyper-Based Rapid Resistance Detection by Stable-Isotope Labeling

Abstract

Against the background of increasing numbers of resistant microorganisms, the fast and cost-efficient detection of microbial resistance is an important clinical requirement for optimal therapeutic intervention. Current routine assays take at least 5 h, but in most cases an overnight incubation is necessary to identify resistant isolates. The usage of matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) profiling in combination with growth media containing isotopically labeled amino acids facilitates the detection of resistant microorganisms after 3 h or less directly from the profile spectrum. Growing microorganisms incorporate isotopically labeled amino acids, increasing protein masses and thereby leading to mass shifts of their corresponding peaks in the profile spectra. In the presence of antibiotics, only resistant microorganisms are able to grow and to incorporate the labeled amino acids. This leads to a difference in the mass spectra of susceptible and resistant isolates, allowing their differentiation. In the presented study, we demonstrated the applicability of this novel approach for the detection of methicillin-resistant Staphylococcus aureus and tested different bioinformatics approaches for automated data interpretation.

INTRODUCTION

The intensive use of antibiotic drugs leads to an increasing number of microorganisms resistant to these antibiotics. Only the specific and restricted use of antibiotics can help to curtail this problem (1). An important prerequisite for the correct choice of the therapeutic intervention is the rapid detection of resistant strains. Common routine methods are based on either automated microbiology systems such as Phoenix (BD Diagnostic Systems, Heidelberg, Germany), MicroScan (Siemens Healthcare Diagnostics GmbH, Eschborn, Germany), or Vitek 2 (bioMérieux, Nürtingen, Germany) (2, 3), which determine the MIC by monitoring the growth of bacteria and are able to provide results after about 5 h for rapidly growing organisms, or agar diffusion assays generally requiring an overnight incubation.

Recently, matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) has become a routine method in microbiology laboratories for the identification of microorganisms (4–6). For example, the comparison of a profile spectrum acquired from a microorganism with corresponding profile signatures stored in a reference library allows for quick and reliable species determination (MALDI Biotyper system; Bruker Daltonik GmbH, Germany). Additionally, MALDI-TOF MS has recently been applied for the detection of ß-lactamase activity providing resistance to β-lactam antibiotics in Gram-negative microorganisms (7–9). This approach, which is a first promising step toward MALDI-TOF resistance detection, monitors the hydrolysis of antibiotics in the presence of the bacteria to be tested. By doing so, it delivers a classification of bacteria according to their ß-lactamase hydrolysis activity within 2 to 3 h (7–9). Unfortunately, this assay is applicable only for detection of resistance to β-lactam antibiotics; in addition, negative results cannot definitely be interpreted as representing susceptibility for all bacteria. Alternative resistance mechanisms might hide the ß-lactamase activity. On the other hand, it can be a rapid and cost-effective tool for observation of the spread of ß-lactamase-carrying bacteria in hygiene monitoring for, e.g., KPC (Klebsiella pneumoniae carbapenemase).

To overcome the restriction of detection of only a single resistance mechanism and to extend mass spectrometric resistance testing to antibiotics other than β-lactams, a novel approach was established. This approach is based on the knowledge that growing cells perform protein biosynthesis. The exchange of naturally occurring amino acids with their nonradioactive isotopically labeled versions in a growth medium leads to the incorporation of these “heavy” amino acids into the newly synthesized proteins. This results in mass shifts of the peaks in the profile mass spectrum of a microorganism. During incubation of bacteria in the presence of antibiotics, resistant strains are still able to grow and to newly synthesize “heavy” proteins. These proteins are detectable as characteristic peak shifts in the profile spectrum. Susceptible strains do not grow or grow only slowly under such conditions, resulting in a profile spectrum most similar to the “normal” profile spectrum of this strain. Thereby, the changes in the profile spectrum should function as an indicator for discriminating between susceptible and resistant strains.

In this paper, the novel MALDI-TOF MS-based resistance test with stable isotopes (MS-RESIST) is described and its applicability for the classification of methicillin-susceptible Staphylococcus aureus (MSSA) and methicillin-resistant Staphylococcus aureus (MRSA) as a model system is demonstrated. First approaches to a bioinformatics data interpretation are shown.

MATERIALS AND METHODS

Bacterial strains and cultivation.

Clinical isolates of Staphylococcus aureus (Max von Pettenkofer-Institute, Munich, Germany) previously classified by standard routine methods as MRSA and MSSA strains were employed as positive and negative samples (10 negative and 10 positive strains), respectively. MRSA strains were characterized as non-borderline oxacillin-resistant Staphylococcus aureus (non-BORSA) strains by detection of the mecA gene and phenotypically by detection of resistance to cefoxitin. Twenty-eight further clinical isolates of Staphylococcus aureus were obtained from the Praxis für Laboratoriumsmedizin, Ärztliche Gemeinschaft für Diagnostik Köln, Bonn, Germany. These were used in a blind manner during the analysis. The molecular characterization is given in the supplemental material. Bacteria were cultivated on Columbia blood agar plates (BD, Germany) without selective agents overnight at 37°C. Fresh overnight cultures were employed for the tests.

Resistance profiling assay.

Dulbecco's modified Eagle medium with low glucose and without lysine, arginine, and leucine (Sigma-Aldrich, Germany) was supplemented with all proteinogenic amino acids (Sigma-Aldrich, Germany) (with the exception of lysine) at a concentration of 0.3 g/liter and with sodium chloride (Sigma-Aldrich, Germany) (4%), glucose (Sigma-Aldrich, Germany) (4 g/liter), and iron-(II) chloride (Sigma-Aldrich, Germany) (60 mg/liter). For each strain, three different 100-μl setups were prepared: “normal,” containing normal lysine (Sigma-Aldrich, Germany) (0.1 g/liter), “heavy,” containing ¹³C₆¹⁵N₂-l-lysine (Fisher Scientific, Germany) (0.1 g/liter), and “heavy + Oxa,” containing ¹³C₆¹⁵N₂-l-lysine (0.1 g/liter) and oxacillin (60 mg/liter) (Sigma-Aldrich, Germany). Each medium preparation was inoculated with bacteria of the same strain at a final concentration of 3.5 × 10⁶ cells per ml. The bacterial suspensions were incubated at 37°C with agitation in a thermomixer (Eppendorf, Germany) for 3 h. After incubation, the cells were centrifuged and washed with 150 μl pure water. Subsequently, bacteria were lysed with 10 μl of 70% formic acid and 10 μl of 100% acetonitrile according to the MALDI Biotyper standard extraction protocol for bacterial profiling (10). Similar experiments were performed with 40 mg/liter cefoxitin (Sigma-Aldrich, Germany) instead of oxacillin.

MALDI-TOF MS analysis.

The cell-free lysates (1 μl) were directly spotted onto a polished steel MALDI target plate. Each lysate was spotted four times. Dried spots were overlaid with MALDI matrix (α-HCCA [10 mg/ml α-cyano-4-hydroxy-cinnamic acid–50% acetonitrile–2.5% trifluoroacetic acid]) (BrukerDaltonik, Bremen, Germany). After drying of the matrix, MALDI-TOF MS measurements were performed with a Microflex LT/SH benchtop mass spectrometer (BrukerDaltonik GmbH, Bremen, Germany) equipped with a 60-Hz nitrogen laser. Parameter settings (ion source 1 [IS1], 20 kV; IS2, 17.5 kV; lens, 6.5 kV; detector gain, 7.4 V; gating, none) had been optimized for the mass range between 2,000 and 20,000 Da. Spectra were recorded in the positive linear mode with the maximum laser frequency. An external standard (bacterial test standard [BTS]; Bruker Daltonik GmbH, Bremen, Germany) was used for instrument calibration.

Data evaluation.

Different data evaluation procedures were applied. First, spectra were visually compared to search for peak shifts using flexAnalysis 3.4 (Bruker Daltonik GmbH, Bremen, Germany) and ClinProTools 3.0 (Bruker Daltonik GmbH, Germany).

Additionally, automated spectra analysis was performed with MALDI Biotyper 3.1 software (Bruker Daltonik GmbH, Bremen, Germany) using the integrated calculation of the correlation index matrix (CCI) and applying the standard settings of the software by considering the mass range between 3,000 to 12,000 Da (correlation analysis). Strains comprising a correlation index for the “normal” spectra versus the “heavy + antibiotic” spectra similar or below 0.6 were classified as resistant. “Heavy” spectra served as a control for growth.

For alternative automated evaluation approaches, specific peaks were chosen for “normal” and “heavy” peaks. The sums of their intensities were changing their ratios during growth in the presence of isotopically labeled amino acids. The following masses were chosen as “normal” peaks: 3,445.1, 4,046.8, 4,307.3, 4,445.3, 4,815.6, 5,033.0, 5,525.5, 6,426.2, 6,892.4, and 9,629.6 Da. The following masses additionally found in the “heavy” spectra were chosen as “heavy” peaks: 3,498.8, 4,076.1, 4,377.2, 4,497.2, 4,864.2, 5,097.4, 5,112.2, 5,613.3, 6,482.2, 6,995.2, and 9,725.2 Da. The ratio of the summarized peak intensities of the “heavy” peaks and the summarized peak intensities of the “normal” peaks [ratio (H/N)] was calculated. This ratio directly correlates with the incorporation of “heavy” amino acids (incorporation analysis). The four different results of each spectra group were displayed in a box plot indicating the median by the bold line, the minimum and the maximum by the whiskers, and the 25th and 75th percentiles by the box. The incorporation analysis resulted in three different H/N ratios for each strain according to each medium.

Considering the ratios of the “heavy + antibiotic” spectra of the incorporation analysis revealed values similar to or greater than 1.0 for resistant strains. For an additional evaluation procedure, the ratios of the incorporation analysis of the “heavy” spectra and the “heavy + antibiotic” spectra were subtracted and subsequently divided by the ratio (H/N) of the “heavy” spectra. This resulted in the normalized difference values. Again, these results were displayed in a box plot. Strains with a normalized difference value similar to or below 0.5 were classified as resistant. The calculations of the incorporation analysis and the normalized difference were performed with a software tool written in the freely available software package “R” (11, 12). All threshold values were empirically set and might be adapted when larger sample numbers have been analyzed.

Routine susceptibility testing of isolates for comparison.

The MICs for oxacillin and cefoxitin were determined, employing the Epsilometer test. Briefly, MIC Test Strips (Liofilchem, Italy) were placed on 85-mm-diameter Mueller-Hinton agar plates (BD Diagnostics Systems) which had been inoculated with a 0.5 McFarland standard suspension of test isolates. All plates were incubated at 37°C for 20 to 24 h before being examined. The MIC was determined to be the value at which the elliptical growth margin intersected the MIC test strip. The evaluation of resistance status was performed according to the current CLSI M100-S21 guideline for aerobic bacteria.

RESULTS

Principle of the assay.

In first experiments, the principle of the mass spectrometric resistance test with stable isotopes (MS-RESIST) assay was investigated. Four known MSSA and four known MRSA strains were incubated for 3 h under the above-described conditions. Each strain was incubated with normal amino acids (“normal”), with isotopically labeled amino acids (“heavy”), or with isotopically labeled amino acids plus antibiotics (“heavy + Oxa”) for 3 h. After the cells had been subjected to washing and lysing, MALDI-TOF MS spectra were acquired. The subsequent spectrum analysis revealed clear differences in the three spectrum sets. Figure 1 shows an enlarged view of representative single spectra shown in Flexanalysis (A and B) and in the “pseudo-gel” views in ClinProTools (C and D) of the three different setups (“normal,” “heavy,” and “heavy + Oxa”) for one of the susceptible strains (A and C) and one of the resistant strains (B and D). For both strains, the “heavy” spectra contained additional peaks not detectable in the “normal” spectra (Fig. 1A and B, boxes). In addition, peaks present in the “normal” spectra were reduced in intensity or even absent in the “heavy” spectra (Fig. 1A and B, gray underlay). Analyzing the “heavy + Oxa” spectra revealed differences in the relative peak intensities of “heavy” to “normal” peaks between the susceptible and the resistant strains. The “heavy + Oxa” spectrum of the resistant strain was nearly identical to the corresponding “heavy” spectrum. This is an indication of growth and protein biosynthesis virtually unaffected by the antibiotic. In contrast, the “heavy + Oxa” spectra of the susceptible strains were more similar to the “normal” spectra than to the “heavy” spectra. This indicates clearly reduced protein synthesis that resulted from the antibiotic action. The ratio of the intensities of the “normal” peaks to the intensities of the “heavy” peaks was quite different in the “heavy + Oxa” spectrum of the susceptible strain. The “normal” peaks in the “heavy + Oxa” spectrum were increased relative to the “normal” peaks in the “heavy” spectrum, and the “heavy” peaks were decreased relative to the “heavy” peaks in the “heavy” spectrum. The display of the spectra in ClinProTools (Fig. 1C and D) allows a direct comparison of the peak intensities because this software normalizes spectra to the total ion count. Additionally, the variance between the different spectra belonging to one set can be easily monitored. In this view, the peaks are displayed as bands and the peak intensities are given as a gray scale. The comparison between the “normal” and the “heavy” spectra shows the maximum change in the peak pattern which resulted from the incorporation of “heavy” amino acids by the growing cells. As expected, no significant differences can be found between the susceptible strain and the resistant strain. In contrast, the “heavy + Oxa” spectra exhibit different patterns for a susceptible strain and a resistant strain. In the case of the resistant strain (Fig. 1D), the “heavy + Oxa” spectra look very similar to the “heavy” spectra. On the other hand, the individual peak pattern of the “heavy + Oxa” spectra of the susceptible strain (Fig. 1C) was still similar to that of the “normal” spectra. In the “heavy + Oxa” spectra, “normal” peaks were more intense than the respective peaks in the “heavy” spectra. In contrast, peaks resulting from the incorporation of “heavy” amino acids were considerably reduced in their intensities compared to the corresponding peak intensities of the “heavy” spectra. The observed relative changes of the peak intensities in the “heavy + Oxa” spectra compared to those of the “heavy” spectra correlated with the reduced growth of the susceptible strains in the presence of the antibiotic and could be used as a measurement to discriminate between susceptible and resistant strains. Using cefoxitin instead of oxacillin led to similar results. Figure 2 represents single spectra from a susceptible and a resistant strain displayed in FlexAnalysis (A and B) and the pseudo-gel views displayed in ClinProTools (C and D).

Fig 1.

Zoom of MALDI-TOF MS spectra displayed in flexAnalysis (A and B) and ClinProTools (C and D) in the mass range between 6,200 and 7,200 Da of a susceptible Staphylococcus aureus strain (A and C) and a resistant Staphylococcus aureus strain (B and D) after incubation with normal lysine (“normal”), with “heavy” lysine (“heavy”), or with “heavy” lysine and oxacillin (“heavy + Oxa”). Peaks corresponding to “normal” proteins are highlighted in light gray (A and B). Peaks corresponding to “heavy” proteins are indicated by the boxes (A and B). y axes give the numbers of multiple measurements (C and D). Intes. [a.u.], intensity [arbitrary units]; Sp.#, spectrum number.

Fig 2.

Zoom of MALDI-TOF MS spectra displayed in flexAnalysis (A and B) and ClinProTools (C and D) in the mass range between 6,200 and 7,200 Da of a susceptible Staphylococcus aureus strain (A and C) and a resistant Staphylococcus aureus strain (B and D) after incubation with normal lysine (“normal”), with “heavy” lysine (“heavy”), or with “heavy” lysine and oxacillin (“heavy + FOX”). Peaks corresponding to “normal” proteins are highlighted in light gray (A and B). Peaks corresponding to “heavy” proteins are indicated by the boxes (A and B). y axes give the numbers of multiple measurements (C and D).

Changes in the peak pattern result in reduced spectrum similarities which can be calculated by correlation analysis and displayed in a composite correlation index (CCI) matrix. Correlation analysis is an established method to compare spectrum similarities. Figure 3 shows the CCI matrices of the susceptible (A) and the resistant (B) Staphylococcus aureus strains analyzed before. For the susceptible strain, a correlation index of about 0.7 was calculated for the correlation between the “heavy” and the “heavy + Oxa” spectra (Fig. 3A). A similar index was calculated for the correlation of “normal” and “heavy + Oxa” spectra. This means that the “heavy + Oxa” spectra represent similarity to the “normal” spectra as well as to the “heavy” spectra. The “heavy + Oxa” spectra seem to represent the averages of the patterns determined for the “normal” and the “heavy” spectra. This result was in concordance with the visual analysis using flexAnalysis and ClinProTools. For the resistant strain (Fig. 3B), the correlation indices were quite different. The correlation between the “heavy” and the “heavy + Oxa” spectra was very strong, represented by a correlation index above 0.9. In contrast, the correlation between the “heavy + Oxa” and the “normal” spectra was quite poor, indicated by a correlation index below 0.6. Again, these results were in good concordance with the visual analysis in flexAnalysis and ClinProTools.

Fig 3.

Correlation analysis: correlation matrix of “normal” versus “heavy + Oxa” versus “heavy” data for a susceptible strain (A) and a resistant strain (B). Red, high similarity; blue, low similarity. Values similar to or below 0.6 for the correlation of “normal” versus “heavy + Oxa” data were considered to represent resistance.

Further bioinformatic methods were developed in order to get defined values which allow a direct comparison of different strains and, thereby, a comparison with the results from routine analyses. The incorporation analysis resulted in three different quotients for each strain (a corresponding figure is shown in Fig. S1 in the supplemental material). The ratio (H/N) for the “normal” spectra was near to zero for all different strains because no “heavy” peaks were present in the spectra derived from growth in normal culture medium. The ratio (H/N) of the “heavy” spectra was a measure for the maximal incorporation of isotopically labeled amino acids and therefore an important control for the growth of the strains. This value differed somewhat for each of the strains regardless of the resistance status. The ratios (H/N) for “heavy + Oxa” spectra revealed clear differences for the susceptible and the resistant strains. The ratios (H/N) were below 1.0 for all susceptible strains. In contrast, the incorporation analysis of the “heavy + Oxa” spectra of the resistant strains revealed values in the same range as the values of “heavy” spectra and clearly exceeded 1.0. In the next step, the normalized differences were calculated for each strain (a corresponding figure is shown in Fig. S2 in the supplemental material). Again, a clear separation between the susceptible and the resistant strains was observed. All susceptible strains revealed a normalized difference above 0.5, and the resistant strains showed normalized differences below 0.5.

Day-to-day reproducibility.

The reproducibility of the mass spectrometric analysis was investigated by analyzing four strains comprising two susceptible and two resistant strains on three different days. The visual inspection of the resulting spectra allowed for a correct classification of the strains according to resistance or susceptibility for all replicates and every day. For an automated classification, the spectrum sets were analyzed with the three different evaluation approaches (Fig. 4). Calculation of the correlation index of the “heavy + Oxa” spectra versus the “normal” spectra revealed only one misclassification. The detailed analysis on the spectrum level showed that these spectra were associated with an increased noise level compared to other spectra. This reduced the similarity to all other spectra, the “heavy” and the “normal” spectra. All other results were in agreement with the expected results (Fig. 4A). The incorporation analysis of the “heavy + Oxa” spectra was in agreement with the expected classification (Fig. 4B). The achieved classification did not agree with the expected one on day 3 for only one of the resistant strains. The detailed analysis of this strain revealed that the ratios (H/N) of the “heavy” and the “heavy + Oxa” spectra were both in the same range but were relatively low. This means that the growth of this strain was poor on day 3. The calculation of the normalized difference revealed complete concordance with the expected classification for all strains on all analysis days (Fig. 4C).

Fig 4.

Reproducibility of the MS-RESIST assay in combination with the different automated data evaluation strategies shown for four different Staphylococcus aureus strains (22318, 55158, 55279, and 56838) and oxacillin. (A) Correlation analysis (“heavy + Oxa” versus “normal”). (B) Incorporation analysis of “heavy + Oxa.” (C) Normalized difference.

Analysis of routine samples.

In the following, the novel approach was applied to routine samples of Staphylococcus aureus and the outcome was compared to the results of the MIC determination. In total, 28 strains were analyzed. Table 1 lists the results of the routine testing and the results of the MALDI-TOF MS analyses for oxacillin and cefoxitin, respectively. For all spectrum sets, the three different evaluation procedures were performed. The results of the automated evaluations (correlation analysis, incorporation analysis of “heavy + antibiotic” spectra, and calculation of the normalized differences) of the mass spectra were mostly in accordance to the routine results. For strain 260, the calculation of the normalized differences resulted in values which were below the threshold values for susceptible strains, resulting in the classification “resistant,” although this strain was classified by other approaches and the routine method as “susceptible.” The detailed analysis of the underlying incorporation analysis revealed a very low value for the “heavy” spectra, representing very poor growth of this strain. The correlation analysis of the cefoxitin approaches resulted in a misclassification in three cases. The detailed visual inspection of the spectra revealed increased noise in the underlying spectra, resulting in decreased similarity to the corresponding “normal” spectra. All other strains were classified in concordance with the different evaluation procedures for the mass spectrometric data and with the routine results determined with all applied bioinformatics algorithms.

Table 1.

Results of resistance testing by the standard routine method and the different bioinformatics approaches applied to the MS-RESIST spectra for the 28 routine isolates^a

Strain	MIC [μg/ml] (R ≥ 4 μg/ml)b		Ratio (H/N)c		Normalized difference of ratio (H/N)		CCI
	Oxacillin	Cefoxitin	“Heavy + Oxa” incorporation analysis (R ≥ 1)	“Heavy + FOX” incorporation analysis (R ≥ 1)	Oxa normalized difference (R ≤ 0.5)	FOX normalized difference (R ≤ 0.6)	“Heavy +Oxa”/“normal” correlation analysis (R ≤ 0.6)	“Heavy + FOX”/“normal” correlation analysis (R ≤ 0.6)
035	256	64	2.15	1.88	0.16	0.12	0.40	0.34
257	256	64	2.59	1.71	0.38	0.47	0.41	0.43
281	4	64	2.43	2.40	0.31	0.06	0.40	0.32
324	0.38	3	0.35	0.42	0.54	0.76	0.73	0.66
330	256	64	1.83	2.08	0.26	0.08	0.40	0.33
331	256	64	1.52	1.88	0.27	0.21	0.45	0.40
384	0.94	3	0.75	0.92	0.76	0.70	0.73	0.69
449	0.125	3	0.77	0.78	0.59	0.73	0.69	0.64
067	32	128	2.40	2.80	0.21	0.09	0.40	0.29
234	2 (4)	8	3.87	2.83	0.17	0.40	0.30	0.32
258	48	64	2.26	1.52	0.48	0.58	0.41	0.50
260	0.38	3	0.89	0.97	0.37	0.23	0.73	0.56
326	0.19	3	0.43	0.60	0.80	0.79	0.73	0.72
595	0.125	3	0.46	0.57	0.77	0.70	0.70	0.70
068	256	64	1.85	2.27	0.19	0.22	0.39	0.35
069	96	64	2.07	3.40	0.18	0.02	0.38	0.29
223	0.19	3	0.73	0.52	0.70	0.86	0.69	0.67
469	0.25	3	0.76	0.52	0.60	0.75	0.74	0.08
496	0.5	3	0.60	0.48	0.73	0.85	0.71	0.68
514	0.25	2	0.79	0.68	0.77	0.73	0.69	0.70
22318	256	64	1.53	2.29	0.20	0.08	0.48	0.05
32241	0.5	3	0.64	0.53	0.80	0.83	0.73	0.02
37828	96	64	1.88	2.91	0.29	0.02	0.39	0.32
55158	256	256	2.46	N/A	0.04	N/A	0.34	N/A
55279	0.125	3	0.89	0.57	0.74	0.77	0.68	0.72
55948	0.25	3	0.62	0.52	0.74	0.79	0.76	0.72
56838	0.125	2	0.56	0.21	0.78	0.79	0.79	0.83
56842	128	64	2.14	1.47	0.39	0.14	0.42	0.41

^aThe resistance (R) breakpoint value is indicated at the top of each column of MIC and MALDI-TOF analysis values. Bold indicates resistance. Italics indicate dissent from other classification results. N/A, spectral acquisition failed.

^bOf the total of 28 strains, 14 were classified as resistant.

^cOf the total of 28 strains, 14, 15, and 14 were classified as resistant by the “heavy + Oxa” incorporation analysis, Oxa normalized difference assay, and “heavy +Oxa”/“normal” correlation analysis, respectively. Of the total of 28 strains, 1 was not measurable and 13, 14, and 16 were classified as resistant by the “heavy + FOX” incorporation analysis, FOX normalized difference assay, and “heavy + FOX”/“normal” correlation analysis, respectively. Oxa, oxacillin; FOX, cefoxin.

DISCUSSION

The principle of the new approach for resistance detection is based on the incorporation of nonradioactive isotopically labeled amino acids into newly synthesized proteins during growth of cells. The use of the isotopically modified proteins led to an increased molecular weight of the peptides and proteins detectable by a mass shift in the MALDI-TOF profile spectra. In the presence of antibiotics, only resistant cells are able to grow and to perform protein biosynthesis. Susceptible strains stop growing under conditions of antibiotic stress and thereby present different profile spectra compared to the setups without antibiotics. Recently, a comparable approach has been published by Demirev et al. (13). In contrast to this study, they employed completely ¹³C-labeled culture medium, resulting in multiple peak shifts with different mass differences which were determined by the number of carbons within the respective peptide and proteins. In addition to the high cost for the culture medium, the evaluation becomes more complicated by this approach.

In contrast, in the study presented here, a protocol and different data evaluation approaches were developed based on the incorporation of single isotopically labeled amino acids. The resulting peak shifts in the profile spectra of Staphylococcus aureus strains were monitored and employed for a subsequent classification.

SILAC (stable isotope labeling by/with amino acids in cell culture) is a common technique for quantitative proteomics (12). Preconfigured cell culture media without certain amino acids, e.g., lysine, arginine, and leucine, are commercially available. For this work, lysine was chosen as the only isotopic amino acid. Lysine occurs frequently in ribosomal proteins, which are the main proteins detected in the MALDI Biotyper profile spectra (14, 15). Employing another amino acid as an isotopic marker led to different changes in the profile spectra (data not shown). To establish the procedure, a small set of Staphylococcus aureus strains with known resistance properties was analyzed with this approach. Different concentrations of “heavy” amino acids were tested (data not shown). The concentration used in this work was a compromise between outcome and costs. Additionally, the incubation time was optimized. An incubation time of 2.5 h already allowed a classification for fast-growing strains but not for all strains. This is attributable to on the one hand a lag phase before the cells start growing and on the other hand the presence of the proteins from the starting material comprising “normal” proteins. Several cell divisions were necessary to reduce the relative amount of the “normal” proteins to get sufficient intensities of the peaks corresponding to the “heavy” proteins. Further, even susceptible strains do not immediately stop producing new proteins in the presence of oxacillin or cefoxitin but show protein production that is reduced over a period of time. Therefore, a sufficient incubation time is required to facilitate the discrimination between susceptible and resistant strains. Thus, the setup with the “heavy” amino acids without antibiotics is an important control because this setup represents the maximal possible growth of a strain. An incubation time of 3 h was found to be sufficient and was applied in all experiments.

The visual inspection on the spectrum level is very time-consuming and not suitable for the analysis of many spectra. Therefore, different automated evaluation approaches were applied to the spectra. Correlation analysis calculates the similarity of the different spectrum sets. A breakpoint had to be defined to get a clear separation between the resistant and the susceptible strains. Since the susceptible strains still perform protein biosynthesis, the “heavy + antibiotic” spectra comprise “normal” as well as “heavy” peaks, resulting in similarities to the “normal” spectra and to the “heavy” spectra. A correlation index below 0.6 for the comparison between “heavy + antibiotic” and “normal” spectra clearly indicated for a resistant strain. Applying this approach to the day-to-day reproducibility test and to 28 unknown samples revealed good concordance with the routine results. In some experiments (Fig. 4C, strain 4, day 3, strain 260/FOX, Strain 469/FOX, strain 32241/FOX), this approach led to wrong classifications. The investigation on the spectrum level revealed that in these cases the quality of the “heavy + antibiotic” spectra was poor. Increased noise resulted in dissimilarity to the “normal” as well as to the “heavy” spectra. This experiment demonstrates the importance of consistent spectrum quality to get correct classifications by correlation analyses.

The incorporation analysis is based on the predefinition of peaks deriving from the “normal” and “heavy” peaks within a spectrum. The ratio of the added intensities of the “heavy” peaks to the added intensities of the “normal” peaks was taken as a measure for the growth. High score values correspond to a growing culture. When susceptible strains were cultured in the presence of antibiotics, this value was decreased compared to that calculated for nonsusceptible strains. A breakpoint of 1.0 for the incorporation of “heavy” lysine in the “heavy + antibiotic” spectra was found to be suitable for the separation of susceptible and resistant strains. Susceptible strains revealed a value below 1.0 for the “heavy + antibiotic” spectra. The evaluation of the spectrum sets according to this method led to the correct classification of all strains but one. In this experiment (Fig. 4A, strain 1, day 3), one of the resistant strains showed a value below the previously defined breakpoint. Investigating this result revealed poor growth of this strain in all setups. Applying the calculation of the normalized difference also revealed good concordance with the routine results. One strain of the unknown clinical isolates was misclassified. The visual inspection of the spectra allowed a clear classification of the isolate as a susceptible strain. The analysis of more strains and strict standardization will be necessary to confirm the suggested breakpoints and to facilitate the handling of strains exhibiting intermediate resistance behavior. In general, the approach presented in this study could be applied to classify Staphylococcus aureus strains according to their resistance status. The use of different antibiotics led to similar classifications. The visual inspection on the spectrum level allowed a correct classification of all strains analyzed, in each experiment. The correlation analysis failed to do so in a few cases. The reasons for the failure were either on the technical side, e.g., the spectrum quality, or on the biological side, e.g., the growth of the microorganisms. Therefore, optimization, adaptation, and further development of the analysis tools will be necessary. Considering that the incorporation analysis and the normalized difference determination are species-specific tools, the development of a species-independent analysis algorithm might be preferred.

Compared to other resistance assays, this approach is very quick. Results could be read out after 3.5 to 4 h, comprising 3 h for the incubation of the cells and the residual time for setting up the assay, the subsequent sample preparation, the acquisition of spectra, and the evaluation. For other microorganisms and antibiotics, even significantly shorter analysis times could be achieved (work in progress; data not shown). Most routine assays need an overnight incubation, and automated resistance detection systems need at least 5 h for fast-growing bacteria before a result can be stated. In laboratories in which MALDI Biotyper-based identification of microorganisms has already been established, no additional technical equipment would be necessary to perform this resistance test. The readout is directly based on the classical profile spectra which are acquired in the same mode as the spectra for species identification. The combination of MALDI-TOF MS identification and MS-RESIST testing as described here might enable identification and resistance testing in a single day on the same equipment. In contrast to the mass spectrometric ß-lactamase (MSBL) assay (7–9), this approach could be applied to different antibiotics additionally to β-lactam antibiotics and might be able to detect all known resistance mechanisms (16, 17). Preliminary results have already demonstrated that this technique can also be applied to investigate meropenem resistance in Klebsiella pneumoniae and Pseudomonas aeruginosa and aminoglycoside resistance in Pseudomonas aeruginosa (data not shown; J. S. Jung, T. Eberl, K. Sparbier, C. Lange, M. Kostrzewa, S. Schubert, and A. Wieser, unpublished data). In principle, the MS-RESIST assay might be generally applicable to different species, including fungi, and to other antibiotics. Optimal conditions for incubation parameters such as time, adaptation of the culture medium, and antibiotic concentration will have to be tested for the respective approaches. Different alternative evaluation procedures might be necessary for optimal results for different species and antibiotic combinations.